The harm lives partly off-platform — how do you measure it without fooling yourself?

Measurement & estimationHard

Problem. For an actor/network harm (say a coordinated influence operation), a lot of the signal isn't in your on-platform impression logs — it's off-platform, and the accounts you can see are the ones your enforcement already caught. How do you build a defensible measurement here?

Before you reveal: say your answer out loud, as if you were in the real interview — get your reasoning across clearly first. There is no single correct answer: reading what the interviewer is really after and defending your own thinking is what makes an answer strong.

← Prevalence is too rare to move in an A/B test — build a metric that can